*OFFICIAL* Alienware m15 R4 / m17 R4 Owner's Lounge

Discussion in '2015+ Alienware 13 / 15 / 17' started by Game7a1, Jan 12, 2021.

  1. Normimb

    Normimb Notebook Evangelist

    Reputations:
    1,354
    Messages:
    453
    Likes Received:
    768
    Trophy Points:
    106
  2. Papusan

    Papusan TURDBOOKs Sucks! Dont waste your $$$ on FILTHY

    Reputations:
    35,908
    Messages:
    27,857
    Likes Received:
    52,946
    Trophy Points:
    931
    It's awful how the boost algorithm works for newer nvidia graphics. It should only throttle down boost clocks in same way as etc Intel processors when it reach TJUNCTION. But I expect its a nice tool to help on RMA costs, as well steal some performance from the users to push them over on new before time. Just sad.
     
    raz8020, etern4l and Normimb like this.
  3. Normimb

    Normimb Notebook Evangelist

    Reputations:
    1,354
    Messages:
    453
    Likes Received:
    768
    Trophy Points:
    106
    Yes it makes a lot of sense. So they don't only cripple performance by removing dynamic boost but as soon as temps are above 40C they reduce clock speed progressively. Nice to know. So do we know for sure what are those decrease values or we user have to guess?
     
    etern4l likes this.
  4. OCTINY

    OCTINY Notebook Guru

    Reputations:
    15
    Messages:
    57
    Likes Received:
    56
    Trophy Points:
    26
    Heh heh, I've been OC'ing for longer than some of you were born :D. Regardless...

    What you're seeing is power throttling, and if you had a 3000 series card, you'd know this. It's not apples to apples w/ Nvidia's previous cards (9 series & up). All 3000 series cards show similar dropped bins, regardless of temps. On average, the 2000 series is much more likely to drop bins due to temp than power, which is the exact opposite of the 3000 series. This is pretty well known in the OC community. And in general, the 3000 series will show higher gains in the 1st test than the 2nd test (which includes much larger drops) vs 2000 series, percentage wise.

    Take any score off the 3000 series Timespy test, (air, chilled, wc'ed etc etc) and this'll prove my point. 99.9% of them show similar drops on the 2nd test, again, regardless of temps. They are power limited before being temp limited, due to how hard they drop and simply the architecture changes in conjection w/ how much power the 2nd test pulls than the first test.

    I can tell you first hand it's not an apples to apples comparison when it comes to overclocking the 3000 series in comparison to the 2000 series. Will bins drop? Absolutely. Is it the prime reason for the percentage drop in the score from the first test temps? Absolutely not.

    Care to explain the dropped bins for the cards below or any 3000 card really on 3Dmarks website, as all show similar drops percentage wise from the first test, regardless of temp. You can't have it both ways.

    https://www.3dmark.com/spy/14869884

    https://www.3dmark.com/spy/18511575

    As I said in my previous posts, what you're saying used to hold true. Not anymore w/ the 3000 series. These are extremely power hungry cards.

    Edit: I don't like coming off as a know it all. It's just I've had enough experience along w/ many others in regards to the 3000 series on overclock.net etc to know. As I said before, it's too bad we won't get a 200w 3080 so it can stretch it's legs a bit more.
     
    Last edited: Feb 24, 2021
  5. Papusan

    Papusan TURDBOOKs Sucks! Dont waste your $$$ on FILTHY

    Reputations:
    35,908
    Messages:
    27,857
    Likes Received:
    52,946
    Trophy Points:
    931
    One of the card stands out. Both for boost, results and temp.
    upload_2021-2-24_6-35-43.png

    All I say is, the cards will drop boost as it will run hotter. Better cooling is equal better performance.
     
    Last edited: Feb 24, 2021
  6. OCTINY

    OCTINY Notebook Guru

    Reputations:
    15
    Messages:
    57
    Likes Received:
    56
    Trophy Points:
    26
    Well it's pretty clear you didn't read my post fully :D

    Why not add the other benchmark 3080 laptop bench I posted? He boosted lower...with colder temps. He's power limited. From the same guy.

    https://www.3dmark.com/spy/18511575

    Also, that 5900hx laptop is vbios limited w/ no mux switch. Now what about the other link w/ the 3090's? And what about the drop in relation to the first test? All similar drops/scores percentage wise, regardless of temp.

    Like I said, all 3000 series cards experience hard drops in the 2nd test due to power than temp. Will you lose some bins due to temps? Of course. But it's power limited more so than ever, especially when it comes to the 1% lows.

    Edit: Nvm, I see you did add it! On my phone :D. But doesn't really help your point.

    Edit 2: And I know that's what you're saying. I'm saying, the drops are bigger due to power limits more so than temp vs 2000 series, on the 2nd test. That's why no matter what 3000 benchmark link you choose, all will have similar drops on the 2nd test, while the 2000 will show less of a drop off on the 2nd test than the first test as it's more temp related vs power related. Along with the architecture changes of the 3000 series, where it gains more in the first test than 2nd test in general, regardless of temp or even power vs 2000 series. That's all I've been trying to communicate since your original comment.

    "This show it crystal clear. Once the GPU test 1 is done, the second graphics test will colapse. Expect same behaviour in games as well."
     
    Last edited: Feb 24, 2021
    Papusan likes this.
  7. etern4l

    etern4l Notebook Virtuoso

    Reputations:
    2,350
    Messages:
    2,734
    Likes Received:
    2,383
    Trophy Points:
    181
    Desktop Ampere cards are power limited? They don't use Max-Q technologies including Dynamic Boost in desktop cards. Strictly speaking there are no Ampere laptop cards on the market, since the new Clevo X170 is not out yet.
     
  8. OCTINY

    OCTINY Notebook Guru

    Reputations:
    15
    Messages:
    57
    Likes Received:
    56
    Trophy Points:
    26
    Yes, specifically in the 2nd test. It hits the vbios limit much quicker in that test than the 1st on a curve in terms of power limit vs temps on all desktop 3000 cards. All well known. These cards while they run cooler on average than the 2000 series, they are starving for power, much more so than the 2000 series ever was. GPU boost will stall due to power limits before temps ever become the *main cuplit*. I've got nearly 500 watts pushing through my shunt modded 3090 to take full advantage of it. Even then it still drops on the 2nd test, due to architecture differences, percentage wise. More so percentage wise then compared to what the 2000 series gets on the 1st test & then on the 2nd test. Once they tap TMSC for the 7nm refresh, I'm sure changes will come back to normalcy & temps will be more of a factor than artificial limits on power via bios (and/or architecture). Or perhaps a driver update will better take advantage of the 2nd test.

    Edit: You do realize the X170 will be using the 165w 3080, right? You will see a similar drop on that system as well percentage wise on the 2nd test, regardless of temps. Scores will be slightly higher due to desktop CPU's in regards to the GPU score overall, but the hard fps drop on the 2nd test will remain. Bank on it. And I'm unsure what you mean they aren't Ampere cards? They are all based on the same architecture. Albiet, different SM and cuda core counts along w/ even worse vbios limits than the desktop versions.

    And when I talk about the Ampere architecture in general performing worse than Turing on the 2nd test, when compared percentage wise on the drops from the 1st to 2nd test to the 2000 series. Look no further. Some top scores for the 3000 series. Temps look great etc

    3060 TI- 40c
    Drop from first test: 15.6%
    https://www.3dmark.com/spy/16054727

    3070- 36c
    Drop from first test: 15.2%
    https://www.3dmark.com/spy/17612119

    3080- 39c
    Drop from first test: 14.6%
    https://www.3dmark.com/spy/18228743

    3090- 14c
    Drop from first test: 14.5%
    https://www.3dmark.com/spy/18352561

    3080 (notebook)- 53c
    Drop from first test: 15.4%
    https://www.3dmark.com/spy/17271051


    2080 Super- 26c
    Drop from first test: 7.3%
    https://www.3dmark.com/spy/18065861

    2080 ti- 19c
    Drop from first test: 8%
    https://www.3dmark.com/spy/17271051

    2080 super (notebook)- 58c
    Drop from first test: 9.5%
    https://www.3dmark.com/spy/18421309
     
    Last edited: Feb 24, 2021
  9. etern4l

    etern4l Notebook Virtuoso

    Reputations:
    2,350
    Messages:
    2,734
    Likes Received:
    2,383
    Trophy Points:
    181
    Sorry, could you clarify how single run scores like the one you posted prove performance drops between runs? I just see single run scores. Even if we did see two results and some slowdowns, we wouldn't be able to tell if this is due to thermal throttling (which is real, particularly on an overclocked 3090) or something else, would we?

    Anyway, stepping back, it is still not obvious by what mechanism would a standard Ampere desktop card power throttle on a second run. The fact that these cards can be shunt modded to use more power is obviously irrelevant.

    Yes, I realize X170 uses the mobile chip... still, it will have the only Ampere mobile card on the market - everything else out there is soldered, although AW takes the infamous Golden Solder prize by also making RAM and WiFi non-upgradable on their new flagship.
     
    Last edited: Feb 24, 2021
  10. OCTINY

    OCTINY Notebook Guru

    Reputations:
    15
    Messages:
    57
    Likes Received:
    56
    Trophy Points:
    26
    Soldiered or not, it's unfortunately a gimped card w/ no pathway to a realistic upgrade (at a price actually worth it). No matter what notebook you choose. Options are great, if they are logical in the long run.

    Wait a sec? Was it not an earlier post that was using a single score run to prove you can expect major thermal throttling in games? You guys can't have it both ways. Want me to gather some high temps score as well? Showing similar differences? I mean I could do this all day :D

    I've shown all the proof that is needed, based on the original comment which used the same premise of a single run. When it comes to the 3000 series, desktop or laptop, you don't base in-game performance off the 2nd test & compare it to the 2000 series, it's not an apples to apples comparison. There's obvious differences which I've gone over countless times. Simple as that. Understandably, it's harder for some to comprehend when most people still can't get their hands on a 3000 series GPU.

    Nevertheless, It definitely looks like I ruffled some feathers here, if anything. So I'll be sure to tune in once that X170 w/ 3080 launches next month ;)
     
    Last edited: Feb 24, 2021
    seanwee likes this.
Loading...

Share This Page